Governments operate large-scale credit assistance programs to support small businesses, yet little is known about their effects on credit allocation, competition, and welfare. We address this gap by studying the SBA 7(a) program, the largest U.S. credit guarantee program, using a model of firm behavior with credit frictions, estimated with data from the hotel industry. We find that 7(a) increases total welfare by $56 million. Of this, $34 million stems from increasing the presence of efficient but credit-constrained firms. The resulting increase in competition lowers prices, generating an additional $22 million. Together, these gains imply a fiscal multiplier of 2.6. We then evaluate alternative program designs. Targeting support toward more concentrated markets generates larger price reductions and raises consumer surplus. A total-surplus-maximizing allocation prioritizes medium-profit firms in larger, less concentrated markets, where lower default risk minimizes fiscal cost per dollar of surplus created. These findings demonstrate that credit frictions can distort market structure, and alleviating them generates welfare gains.

To match highly volatile demand with fixed capacity, cloud computing providers employ tiered reliability---offering users discounted spot compute from which they can be evicted with little warning when capacity tightens. We study this design using proprietary data from a major cloud platform combining VM-level usage, prices, and evictions, together with a randomized price experiment. Reduced-form evidence shows: (i) demand is price-responsive, with an average own-price elasticity around -0.5; (ii) evictions persistently reduce subsequent usage by 40%, indicating a strong revealed preference for reliability; and (iii) additional aggregate usage increases eviction rates for spot workloads, consistent with congestion as the system approaches capacity. We interpret these facts through a demand model where heterogeneous users choose the level of compute reliability for each workload, while learning about eviction risk through experience. On the supply side, evictions arise endogenously as a function of usage given fixed capacity constraints. Our results quantify how spot compute exploits heterogeneity in reliability preferences to utilize idle capacity while concentrating eviction costs on users who value reliability the least.

Advertising is a central marketing tool, yet credible evidence on its direct effect on sales is limited. We propose an additional motive for heavy advertising in concentrated markets: entry deterrence. Using a merged dataset of U.S. television advertising and sales in the chocolate industry, we document reduced entry in categories with intensive advertising. Exploiting quasi-experimental variation from coarse spatial targeting, our reduced-form estimates show that (i) the short-run impact of a brand’s own ads on its sales is modest, but (ii) a 10% increase in rivals’ total advertising lowers a brand’s sales by 1.6%. Motivated by these facts, we develop a dynamic model of product entry and advertising that captures entry deterrence through advertisers' ability to reduce profits of potential entrants. We will estimate the model to quantify the share of advertising attributable to entry-deterrence motives and to evaluate counterfactual policies that limit advertising intensity and their welfare implications.